OpenAI gives safety board power to delay launches

Putting the brakes on frontier AI has always sounded like wishful thinking, especially when the people hitting pause are the same ones racing to ship.

OpenAI just turned its Safety and Security Committee into an independent board oversight body. The key takeaway: the committee now has the formal authority to delay a model launch until safety concerns are resolved.

Who is holding the emergency brake

The move follows a 90-day review of OpenAI's internal safeguards.

Heading up the newly independent committee is Zico Kolter, director of Carnegie Mellon University's machine learning department.

He is joined by Quora CEO Adam D’Angelo, retired US Army General Paul Nakasone, and Nicole Seligman, former executive vice president and general counsel at Sony.

Unlike its previous setup, which included company executives like Sam Altman, this refreshed group consists strictly of independent board members.

OpenAI confirmed that this committee already vetted the safety criteria and evaluation results for the recently released o1 model prior to launch.

Fortress security and outside eyes

Beyond boardroom governance, OpenAI outlined several concrete operational changes.

On the cybersecurity front, the company is rolling out stricter internal information segmentation, expanding round-the-clock security operations, and exploring an industry-wide Information Sharing and Analysis Center for pooling cyber threat intelligence.

They are also bringing in outside inspectors. That includes partnering with Los Alamos National Laboratory to study AI risks in bioscientific research, alongside formal pacts with both the US and UK AI Safety Institutes.

Internally, OpenAI restructured its research, safety, and policy teams to unify safety protocols across the entire company.

Why it matters

Separating safety oversight from executive leadership is a smart and necessary move. Historically, lab leaders have graded their own homework, creating an inevitable conflict between commercial release dates and precautionary guardrails.

Giving independent directors the explicit power to halt a model deployment gives OpenAI a much stronger defense against regulatory scrutiny and public skepticism.

However, the real test will come down to enforcement. Having the theoretical authority to delay a rollout is one thing; actually pulling the plug on a multi-billion-dollar frontier model release while rivals are breathing down your neck is another matter entirely.